This article presents a model of the research and publication process that analyzes why scientists commit fraud and how fraud can be detected and prevented. In the model, authors are asymmetrically informed about the success of their projects and can fraudulently manipulate their results. We show, first, that the types of scientific frauds that are observed are unlikely to be representative of the overall amount of malfeasance; also, star scientists are more likely to misbehave but less likely to be caught than average scientists. Second, a reduction in fraud verification costs may not lead to a reduction of misconduct episodes but rather to a change in the type of research that is performed. Third, a strong "publish or perish" pressure may reduce, and not increase, scientific misconduct because it motivates more scrutiny. Finally, a more active role of editors in checking for misconduct does not always provide additional deterrence. The Author 2009. Published by Oxford University Press on behalf of Yale University. All rights reserved. For Permissions, please email: journals.permissions@oxfordjournals.org, Oxford University Press.